|
⏱ The 10-Second Version
- Anthropic began embedding an invisible watermark in everything Claude writes, complying with the EU AI Act’s Transparency Code, and is building tools so outsiders can check for it themselves.
- Nvidia partnered with six Wall Street giants, Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR, to mobilize over $500 billion turning AI chips into a tradeable asset class.
- Mark Zuckerberg’s new essay describes a “personal superintelligence” that watches him sleep, coaches his workouts, and bakes with his 8-year-old daughter, drawing a “dystopian” label from an MIT researcher.
|
|
The Big Thing
Anthropic Starts Invisibly Watermarking Every Word Claude Writes
What Happened
Anthropic confirmed this week that every Claude model released since August 2 embeds an invisible watermark in its text, fulfilling its obligations under the EU AI Act’s Transparency Code, which took effect that same day. The mark survives copy-paste and some editing, though heavy rewriting, paraphrasing, or translation can erase it; Anthropic is building tools to let outsiders check for it themselves. The timing is pointed: Hachette recently pulled a novel over suspected AI authorship, and a literary agent dropped another book on the same grounds, exactly the proof publishers are now hunting for. Read the full story at TechCrunch →
Why You Care
Ask your team this week whether anything shipped under your name was drafted by Claude, and whether that would matter if it came out.
|
What to Say in Your Next Meeting
“Anthropic just started invisibly watermarking everything Claude writes, so text you paste into your own work now carries an AI fingerprint you didn’t put there.”
|
|
|
Speed Round
Nvidia partnered with six Wall Street giants, Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR, to establish financing platforms aiming to mobilize more than $500 billion in third-party capital for AI data centers, explicitly framing Nvidia compute as “an investable asset class”; CEO Jensen Huang put it plainly: “In AI, compute is revenue.” Turning chips into a $500 billion asset class is either brilliant financial engineering or a plot Wall Street has run before.
Read the announcement at Nvidia →
Anthropic partnered with Macquarie Asset Management and Singapore’s GIC to launch Theseus Infrastructure, a data center platform where Anthropic anchors long-term leases while Macquarie and GIC fund and own the majority of each project; Anthropic also committed to cover any electricity price increases nearby consumers face because of the buildout. Promising to eat your neighbor’s power bill is a new one, even for an industry used to promising almost anything to get a data center built.
Read the announcement at Macquarie →
Intel is selling $15 billion in new stock, about 3% dilution, to fund AI chip and foundry expansion, telling investors demand is “outpacing the company’s growing supply” of wafers, memory, and substrates; shares fell nearly 5% in premarket trading anyway. Intel raised $15 billion to chase the AI boom, and investors’ first reaction was to sell.
Read the full story at SiliconANGLE →
Mark Zuckerberg published an essay describing his vision of “personal superintelligence,” a 24/7 AI agent that tracks his sleep, coaches his workouts, and bakes with his 8-year-old daughter each weekend; MIT’s Sherry Turkle called the vision “dystopian,” and critics noted Meta is pitching always-on family monitoring months after pausing teen AI-character access and while facing an FTC probe into child chatbot safety. Nothing says “the future is personal” like an AI that watches your kid bake and reports back.
Read the full story at Fortune →
|
|
Prompt of the Day
The AI Disclosure Line
Anthropic just made your Claude drafts individually traceable, whether you planned for that or not. Most teams have never actually decided what “used AI to help write this” means for their own work, so the watermark ends up deciding for them by default.
I use Claude (or another AI tool) to help draft [paste: e.g. emails, reports, code, articles, proposals] for my job.
Help me write a short, honest disclosure policy that covers:
1. Which of these I should flag as AI-assisted, and to whom
2. What "AI-assisted" actually means once I've heavily edited the output myself
3. One sentence I could say out loud if a client or manager asked, without sounding defensive
Keep it under 100 words so I'll actually use it.
|
|
Forward this to the coworker who’s never once mentioned that the email actually came from Claude.
— Don. Ran this on my own drafting process writing this issue. Turns out I already say “written with Claude” more than I say “written by me.”
|
